An analysis of coincidence detector networks

A coincidence detector is a neuron model which has a type of dendrites whose time constant is small, then a single pulse is more significant than the case of integrator neurons and the neuron can detect synchronization of pulses from distinct connections. Our aim of this study is to investigate the characteristics not of a single or a few neurons, but of larger networks with the coincidence detectors. Networks responding to one or two sequences of pulses are constructed by selecting suitable connections in random-connected networks. It is shown that the selected networks can response-only to a specific sequence of pulses even if noisy pulses are superimposed on it. It is also shown that the activation of some neurons in the networks can be partitioned for an alternative sequence by observing cross-correlations of output pulses, when two sequences of pulses are supplied to the networks simultaneously.